Automated Insurance claims processing using Amazon Bedrock, Knowledge Base and Agents
A solution to automate the insurance claims processing lifecycle
Majid Shokrolahi
Amazon Employee
Published May 8, 2024
In the dynamic insurance industry, companies are constantly seeking ways to streamline their operations and deliver exceptional customer experiences. One such approach is the integration of various AWS services to automate the insurance claims processing lifecycle. In this blog, we'll explore how to leverage the capabilities of Generative AI along with AWS Serverless technologies, including Amazon Bedrock, to operate Large Language Models (LLMs) while ensuring data security and privacy. This technical article will explore the AWS architecture and the key services that enable this powerful automation. For those interested in a demo and deployment solution with deeper insights and explore how the solutions works refer to this Github repository.
Amazon Bedrock provides a scalable platform for building AI-powered applications. It is a managed service that enables customers to deploy and use large language models securely. Bedrock ensures customer data is kept private and is not used to train or fine-tune the base models, protecting the confidentiality and integrity of sensitive information. It's designed to facilitate dynamic, context-aware responses to customer inquiries, simulating human-like interactions that offer real-time solutions and advice. By analyzing large volumes of data, Bedrock can provide personalized service options based on individual customer profiles and historical data.
A key feature of Amazon Bedrock is its ability to create and manage extensive Knowledge Bases. Knowledge bases provide fully-managed RAG to supply the agent with access to your data. These knowledge bases are crucial for understanding complex insurance policies and regulations. They allow generative AI systems to access a vast array of structured information, enabling them to provide informed responses and make intelligent decisions during the claims process.
Front-end Integration:
The process begins with the mobile client, where customers can initiate insurance claims. The AWS SDK enables seamless integration between the client and the AWS services, ensuring secure and efficient data exchange. A typical claim would contain a filled form and, optionally, images to accompany the claim.
The process begins with the mobile client, where customers can initiate insurance claims. The AWS SDK enables seamless integration between the client and the AWS services, ensuring secure and efficient data exchange. A typical claim would contain a filled form and, optionally, images to accompany the claim.
Authentication and Authorization:
Amazon Cognito plays a crucial role in managing user authentication and authorization, ensuring that only authorized individuals can access and interact with the claims processing system.
Amazon Cognito plays a crucial role in managing user authentication and authorization, ensuring that only authorized individuals can access and interact with the claims processing system.
(Optional) Images are stored on Amazon S3:
If any images are attached to the claim, they are securely uploaded into an encrypted S3 bucket. You can point the knowledge base to your Amazon S3 data source.
If any images are attached to the claim, they are securely uploaded into an encrypted S3 bucket. You can point the knowledge base to your Amazon S3 data source.
Initiate claim processing:
Mobile application initiates claim processing by sending the request to backend using Amazon API Gateway.
Mobile application initiates claim processing by sending the request to backend using Amazon API Gateway.
Step 1. Claim processing job
As claim processing job requires several steps performed by multiple backend components, it will benefit from the job orchestrator. The AWS Step Functions service coordinates parallel processes, exception handling, retries, and timeouts based on the specified business logic, eliminating the need to manually orchestrate application components. It also automatically handles errors, and restarts to ensure that application tasks are executed as expected, reducing the number of failed user requests.
As claim processing job requires several steps performed by multiple backend components, it will benefit from the job orchestrator. The AWS Step Functions service coordinates parallel processes, exception handling, retries, and timeouts based on the specified business logic, eliminating the need to manually orchestrate application components. It also automatically handles errors, and restarts to ensure that application tasks are executed as expected, reducing the number of failed user requests.
Storing the claim
Submitted form and links to the images are stored in the data warehouse or operational claims database.
Submitted form and links to the images are stored in the data warehouse or operational claims database.
Step 2. Intelligent Claims Processing using Amazon Bedrock
At the heart of this automated insurance claims processing solution lies the integration of Amazon Bedrock, a powerful platform that enables the deployment of large language models and generative AI capabilities.
Amazon Bedrock is a fully managed service that provides access to a variety of high-performing foundation models (FMs) from leading AI companies like Anthropic, Cohere, Meta and Mistral AI. This allows customers to easily experiment with and evaluate different models for their use cases without exposing your data to train the underlying models. Multi-modal foundational models such as Anthropic Claude 3 Sonnet are especially beneficial for claims processing as they are able to accept both text and images data as an input and handle image classification and understanding tasks.
Amazon Bedrock is a fully managed service that provides access to a variety of high-performing foundation models (FMs) from leading AI companies like Anthropic, Cohere, Meta and Mistral AI. This allows customers to easily experiment with and evaluate different models for their use cases without exposing your data to train the underlying models. Multi-modal foundational models such as Anthropic Claude 3 Sonnet are especially beneficial for claims processing as they are able to accept both text and images data as an input and handle image classification and understanding tasks.
Additionally, Bedrock offers a Retrieval Augmented Generation (RAG) feature, which integrates the foundation models with internal data sources through Amazon Bedrock Knowledge Bases, making the responses more contextual and accurate. For example, using Amazon Bedrock Knowledge Base to store insurance policy documents allows for automated checks of the claim against the policy, including the specific policy excerpts to explain the decision.
Amazon Bedrock also includes Agents, which are generative AI programs that can automate multi-step tasks by orchestrating actions, using knowledge bases, and generating responses based on user queries. For claims processing, in addition to policy validation, more information about the customer is required to asses the claim, such as history of previous claims or customer details.
To ensure responsible AI development, Amazon Bedrock provides Guardrails, which allow configuring rules to control denied topics, content filtering, and privacy protection, aligning the AI applications with organizational policies and ethical standards.
Amazon Bedrock helps ensure that your data stays under your control. When you tune a foundation model, we base it on a private copy of that model. This means your data is not shared with model providers, and is not used to improve the base models. You can use AWS PrivateLink to establish private connectivity from your Amazon Virtual Private Cloud (VPC) to Amazon Bedrock, without having to expose your VPC to internet traffic. Finally, Bedrock is in scope for common compliance standards including ISO, SOC, CSA STAR Level 2, is HIPAA eligible, and customers can use Bedrock in compliance with the GDPR.
Using a combination of foundation models, Knowledge Base and agents a claim assessment report is automatically generated and can be reviewed by human subject matter experts.
Step 3. Human validation
AI models, while highly capable, are still evolving to match human accuracy can still produce inaccurate, biased, or unreliable content. Human subject matter experts can review the generated claim assessment report to ensure it aligns with facts, company policies and ethical guidelines and make necessary adjustments if required.
The integration of AWS services, including Amazon Bedrock, Knowledge Bases, and Intelligent Agents, unlocks numerous benefits for insurance companies:
- Increased Efficiency: Automation reduces the time and effort required to process claims, improving accuracy and reducing error rates.
- Enhanced Customer Experience: Faster claims processing and personalized responses lead to higher customer satisfaction and loyalty.
- Operational efficiency: Automation decreases the manual effort involved in claims handling, allowing claims handlers to focus on more complex cases. In addition, AWS Serverless technologies enable automatic scaling to adjust resource use dynamically. This optimizes expenses and enhances developer productivity by allowing them to focus on adding business value rather than infrastructure management.
- Proactive Regulatory Compliance: The system can automatically detect and implement the latest legislative changes, ensuring the company's policies and procedures are up-to-date. For instance, when new legislation is passed that affects insurance operations, Amazon Bedrock can detect these changes through its automated scanning of regulatory websites or a data source. Once a new regulation is identified, it could notify it for a human review and then the system can seamlessly integrate this change into the existing operational framework of the company. It updates the knowledge bases that guide the claims processing, inform staff on the new claim requirements, and modifies workflows to adhere to the updated legal standards. This proactive approach not only ensures compliance but also enhances the company's agility in adapting to regulatory changes, thereby protecting the company against potential legal challenges and fines. This automated compliance is a significant advantage for insurance companies, enabling them to stay current with the latest regulations without dedicating extensive resources to manual compliance efforts. This efficiency also helps in maintaining trust and reliability in the eyes of customers and regulatory bodies.
- Improved Claims Forecasting: Predictive modeling and data-driven insights enable better resource planning and decision-making.
In this post, we explored a solution to automate the insurance claims processing lifecycle and how Amazon Bedrock could help insurance companies through a strategic shift towards advanced automation. Amazon Bedrock, alongside Knowledge Bases and Intelligent Agents, seamlessly manages the claims lifecycle from start to end. Knowledge Bases store up-to-date policies, enabling Agents to extract data, make informed decisions, and deliver accurate, personalized responses to customers, thereby enhancing efficiency and satisfaction. The solution and building blocks can be found and deployed as part of our GitHub repo, or adjusted for your specific use-case needs and model choice.
For businesses, this translates into reduced operational costs, optimized resource allocation, and an elevated customer experience—key drivers for competitive advantage in today's rapidly evolving insurance market.
For businesses, this translates into reduced operational costs, optimized resource allocation, and an elevated customer experience—key drivers for competitive advantage in today's rapidly evolving insurance market.
Majid Shokrolahi is Senior Solutions Architect at AWS, focused on Italian Startups helping them to innovate and build their solutions on the AWS platform. He is passionate about Containers , Gen AI and the startup ecosystem.
Alex Tarasov is Senior Solutions Architect working with Fintech Startup customers helping them to design and run their data and AI/ML workloads on AWS. He is a former data engineer and is passionate about all things data.
Any opinions in this post are those of the individual author and may not reflect the opinions of AWS.